Machine Learning models present a transformative method of optimization and prediction in science and engineering research. In the chemical sciences, unsupervised deep learning models such as autoencoders have shown to be useful for property prediction and material discovery. One area of interest is for the discovery of high performing electrolytes with wider electrochemical stability for increased energy density of batteries. The problem with current machine learning models is that they require an abundance of chemical property data. To help address this question we present an online simulation tool hosted on nanoHUB that allows the user to joint train a model on two disparate data sets. This model allows the end-user to increase their knowledge on a scarce data set by using a data-rich property set. We also investigate the effect of chemical representation and autoencoder type on property prediction and compound generation.

Bryan Arciniega is a third-year undergraduate at California State Polytechnic University, Pomona who is studying computer engineering and finance. He currently works as an IT technician for Cal Poly?s Student Health Services where he helps to secure and scale the technology infrastructure of the firm. He is also a part of the Student Managed Investment Fund where he uses his quantitative skills to invest Cal Poly Pomona's assets. His prior research investigates machine learning techniques and how they can be implemented to build financial models. In the summer of 2019, he worked with the Network for Computational Nanotechnology (NCN) at Purdue University under the mentorship of the Savoie group. He plans to receive his Master?s in Financial Engineering where he can learn to leverage, finance, mathematics, and computer programming to make portfolio management decisions.

NCN Undergraduate Research Experience 2019

Researchers should cite this work as follows:

• Bryan Arciniega (2019), "3 min Research Talk: Web-based Machine Learning Tool for Material Discovery and Property Prediction," https://nanohub.org/resources/31336.

  BibTex | EndNote

1. chemistry
2. machine learning
3. material properties
4. materials science
5. ncndev
6. undergraduate research